Translational Evaluation of a Machine Learning-Based Interactive Lab for Aphasia Rehabilitation in Post Stroke...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Rehabilitation Medicine

Background:

• Conventional aphasia therapy has limitations in personalization and scalability.
• Post-stroke aphasia significantly impacts patients' quality of life and communication abilities.
• There is a need for innovative, data-driven approaches to aphasia rehabilitation.

Purpose of the Study:

• To develop and clinically evaluate a machine learning-based interactive lab for personalized aphasia rehabilitation.
• To assess the efficacy of the Language Interactive Lab compared to conventional therapy.
• To leverage machine learning for tracking aphasia severity and recovery.

Main Methods:

• A four-week randomized clinical trial involving 27 aphasia patients.
• Experimental group used the Language Interactive Lab; control group received conventional therapy.
• Language performance assessed via the Chinese Communicative Aphasia Test (CCAT); system data used for ML classifiers.

Main Results:

• The experimental group showed statistically significant improvements in 7 out of 9 CCAT subtests and overall score.
• Machine learning classifiers achieved up to 91.7% accuracy in predicting aphasia severity and recovery.
• The interactive lab demonstrated clinical efficacy in enhancing language outcomes.

Conclusions:

• The interactive lab integrates gamified therapy with real-time ML assessment for effective aphasia rehabilitation.
• The system offers a scalable framework for AI-driven, adaptive neurorehabilitation.
• Clinically validated and designed for regulatory compliance (TFDA SaMD), enabling translational deployment.